Mechanical fault diagnosis method of self-supervised convolutional neural network

• Main Authors: HU CHEN, ZHANG ZHIDONG, LI WENCHAO, WAN FU, ZHANG JIAN, MAO HU, SHEN XURUI, ZHOU YONGLIN, WANG HAO, WANG FENGYU, GAO YUAN, WANG WENQUAN, SONG XIN, LIAO FEILONG, YU JIANSHENG, XU YOUHONG, QIN LIU
• Format: Patent
• Language: Chinese; English
• Published: 12.03.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention relates to the technical field of mechanical fault diagnosis, in particular to a mechanical fault diagnosis method of a self-supervised convolutional neural network, which specifically comprises the following steps: setting a signal transformation-based auxiliary task of a classification task type, inputting an unlabeled signal and endowing the unlabeled signal with a pseudo label corresponding to a signal transformation method; taking the generated data set with the pseudo label as a training sample to train a neural network in a mechanical fault diagnosis framework until convergence, freezing part of preposed convolutional layer parameters of the neural network, migrating the neural network to a fault diagnosis task, performing supervised learning on a limited number of obtained labeled data sets, and obtaining a fault diagnosis result; and updating unfrozen parameters of the neural network, and finally obtaining a mechanical fault diagnosis model. Through the diagnosis method, the problem tha